Six months of published research, read together, describe a recovery environment harder than the one most 2026 budgets were built for.
Four findings carry the weight. Volumes rose while the money got harder to collect, denial economics deteriorated at both ends, regulatory pressure shifted off the federal enforcer and onto states and courts, and the risk in outsourced work concentrated in the vendor layer.
Every number below comes from a named organization’s published research. Where we use one of our own figures, we say so.
Volumes rose while liquidity fell
TransUnion’s seventh annual Debt Collection Industry Report: Investing For Impact, found 64% of collections operations saw account volumes increase in 2025. 75% expect more volume over the next twelve months, and 48% forecast double-digit growth.
The same survey found 34% of third-party agencies reporting decreased account liquidity, and 28% reporting declining collectability.
More accounts arriving, each one worth less. That combination breaks the capacity math most recovery operations run on, because staffing models assume a stable relationship between placement counts and revenue.
The Federal Reserve Bank of New York’s Q1 2026 household debt report shows where the volume comes from. US household debt reached $18.794 trillion, with 4.8% in some stage of delinquency. Serious delinquency transitions on student loans hit 10.86%, up from 8.04% a year earlier, and roughly 2.6 million borrowers more than 120 days past due were transferred to the Department of Education’s Default Resolution Group. Credit card serious delinquency transitions sit at 7.10%.
One counter-signal is worth holding onto. The Federal Reserve Board’s credit card charge-off rate fell to 3.84% in Q1 2026 from 4.46% a year earlier, its fifth consecutive quarterly decline. Charge-offs and delinquency transitions pointing in opposite directions usually means issuers are holding and working paper longer before it leaves the balance sheet.
Both readings point the same way for planning. Build the second half on yield per account, and shorten the time paper sits before placement. The Commercial Collection Agencies of America collectability benchmark, republished by the Credit Research Foundation, puts a receivable at 88.7% collectable one month past due and 51.3% at six months. That’s 37 points lost to the calendar. No fee negotiation in the history of receivables has moved 37 points.
Denial economics got worse at both ends
Kodiak Solutions’ March 2026 benchmarking, covering roughly 2,300 hospitals, found provider organizations lost $48.4 billion in net revenue to final denials and bad debt in 2025, up 25% from $38.6 billion the year before. The average initial denial rate rose to 11.6%. The rate at which providers overturn initial denials fell to 42.1%.
Both ends moved the wrong way in the same year. More claims denied, a smaller share recovered.
Kodiak also found patient responsibility rose from 6.8% to 7.3% of net patient revenue while the collection rate on that patient balance fell from 45.1% to 42.4%. More of the bill sits with the patient, and less of it gets paid.
What makes this expensive rather than merely irritating is how much of the denied volume was payable all along. Premier’s survey of 280 hospitals across 23 states, covering 2023 claims, found roughly 70% of initially denied claims eventually get paid, against administrative expense per claim that rose from $43.84 to $57.23. Premier put total claims adjudication cost for that year at $25.7 billion, with about $18 billion of it potentially unnecessary.
The appeal is the least-used instrument on the floor. KFF’s March 2026 analysis of ACA marketplace plans found fewer than 1% of denied claims are ever appealed, and insurers upheld the original denial in 66% of the appeals they did receive. A JAMA study of roughly 51,000 New York external appeal cases, published 15 April 2026, found the share of denials overturned on appeal rose from 38% in 2019 to nearly 53% in 2025, with home health reversals above 78%.
Appeal capacity is the binding constraint here, and it’s a staffing problem before it’s a technology problem. Experian Health’s 2025 survey found 90% of denials require at least some human rework, and 68% of providers say submitting clean claims is harder than it was a year earlier.
One figure of ours, marked as ours. PULSE, our denial analysis and appeal drafting system, runs an 86% overturn rate on the denial populations we work, with 43% higher payment per appeal and half the time to appeal. Results vary by payer mix and client program. The source is our own operating data, measured internally rather than audited by a third party, and it isn’t a prediction of what any other organization would see.
The regulatory center of gravity moved to states and courts
The federal collections regulator stepped back in a way that’s easy to misread as relief.
By August 2025 the CFPB had withdrawn 67 guidance documents, per GAO, including its circulars on algorithmic credit decisions and three debt collection advisory opinions. It dismissed 19 of the 34 enforcement actions live in January 2025. Its own November 2025 FDCPA annual report shows no federal FDCPA enforcement action brought in 2024. The 2026 regulatory agenda carries one debt collection item, a definitional question about larger participants.
What filled the space is harder to plan around than a single rulebook. NYU’s Center on Technology Policy counted 109 state AI laws enacted across 29 states in the first six months of 2026. Texas TRAIGA took effect 1 January 2026 with penalties of $80,000 to $200,000 for uncurable violations. California’s ADMT regulations require compliance by 1 January 2027. Colorado repealed its 2024 AI Act in May 2026 and replaced it with a disclosure, notice, and human-review statute effective January 2027. Holland & Knight counted six states enacting laws in 2026 saying an insurer can’t rely on AI as the sole basis for denying care.
Private litigation moved the same direction. WebRecon’s 2025 counts show FCRA filings up 37.4%, FDCPA filings up 7.8%, and CFPB complaints up 89.1% year over year. The trade tracker TCPAWorld reported 330 TCPA cases filed in April 2026 alone, 255 of them class actions, up 40% on the prior year (single-source, directional).
Exposure repriced while the rulebook stayed still. FDCPA class damages are capped at the lesser of $500,000 or 1% of net worth. TCPA damages run $500 per violation, trebled for willful conduct, with no aggregate cap. Any operator adding automated voice or messaging is picking up telecom-statute risk, and most haven’t repriced for it.
Third-party risk concentrated in the seam
Verizon’s 2026 Data Breach Investigations Report found 48% of confirmed breaches involved a third party, up 60% year over year. Across three consecutive editions the figure ran 15%, then 30%, then 48%.
The healthcare version is sharper. HIPAA Journal’s June 2026 analysis of HHS Office for Civil Rights breach portal data found that in 2015, 5% of individuals exposed in reported healthcare breaches were exposed through a business associate. In 2025, 65% were. Business associates were involved in 43% of reported healthcare breaches in the first half of 2026.
The Identity Theft Resource Center’s H1 2026 report shows the mechanism: 38 supply chain attacks produced 280.6 million victim notices across 206 organizations. Roughly five organizations compromised per incident.
The failure point behind each of those numbers is the handoff. Data crossing an organizational boundary under mismatched controls, with nobody accountable for the whole path, is where recovery work breaks. Reducing vendor count is one way to remove those handoffs, and it buys concentration risk in exchange. We’ve argued the honest version of that trade in The seam is the failure point.
Four things worth doing in the second half
Place earlier. The decay curve costs more than any rate negotiation returns, and placement timing is usually set by fiscal calendar rather than by yield.
Fund the appeal function before buying more appeal software. Experian Health’s rework finding says the constraint is people who can work a denial, and the overturn data says a lot of denied claims were payable all along.
Inventory your models and your consent records now, while it’s a project instead of a discovery response. Every state regime that landed in 2026 is satisfied by producing records.
Map your seams. Count how many organizational boundaries a single account or claim crosses between placement and resolution, name who owns the outcome at each one, and keep the map current as the panel changes.
TSI applies AI and automation across recovery and revenue cycle work, including predictive account scoring, channel orchestration, real-time agent assistance, automated quality monitoring, and generative drafting for denial appeals, inside an audited compliance management system with human review retained on consumer-facing and clinical-adjacent decisions.
Sources: TransUnion, seventh annual Debt Collection Industry Report; Federal Reserve Bank of New York Q1 2026 Household Debt and Credit; Federal Reserve Board via FRED; Commercial Collection Agencies of America via Credit Research Foundation; Kodiak Solutions (March 2026); Premier Inc.; KFF (March 2026); JAMA (April 2026); Experian Health State of Claims 2025; GAO-26-108448; CFPB FDCPA Annual Report (Nov 2025); NYU Center on Technology Policy; Holland & Knight; WebRecon via Troutman Pepper Locke; TCPAWorld; Verizon 2026 DBIR; HIPAA Journal analysis of HHS OCR data (June 2026); Identity Theft Resource Center H1 2026.